Chip operating temperatures

[tom cloud]

Oh me, PCB's are *extremely* durable!  You can often bend one almost
double without causing any damage.  Hair line cracks are usually
caused in the manufacture process due to a slight imperfection
in either the mask (negative or positive) or the resist.  1/16"
epoxy-glass boards are made from (typically) four layers of fine
fiberglass cloth impregnated with epoxy resin.  They last for
many years -- there's stuff out there that's been working everyday
for over twenty years!!!  The major problem in circuit board failure
(and this is an industry statistic) is solder joints.  The major cause
of problems in circuits overall is connectors.  The rule of thumb
is solder it rather than use a connector (includes IC sockets) unless
it has to be removable.  It is not true that the boards dry out or
get brittle -- not epoxy glass.  Phenolic starts out brittle (that's
what you typically get if you buy from Radio Shack!) and is a very poor
substrate (that's what you call the stuff under the copper foil
the traces are made from).  The wiring in car dashes is usually
a flexible PCB made from a mylar film with the copper foil on it.

It is true that problems can be created by the way components are
mounted on the PCB.  For example,  GE made a two-way radio still
used in many police, taxi, and other commercial apps called a
Master-II.  The audio board is surrounded by a casting and is
about 5" square (been several years since I've been into one).
The push-pull audio output stage was two TO-220 transistors (much
like the transistors used to pulse injectors in EFI circuits)
which were soldered to the board and bolted to the casting for
heat dissipation.  A common occurence was intermittent or failed
audio due to the leads pulling away from the transistor's chip inside
its header.  See, the board was firmly mounted and the transistor
was also, being bolted down, so as it heated and cooled there
was no room for expansion/contraction eventually causing failure.
This problem was easily fixed by putting a little "jog" in the
transistor's leads so it could move a teensy bit.  (Yes, "teensy"
*IS* a technical term!)

I used to work with shake tables big enough to put a refrigerator on.
We simulated the vibes from rocket engines.  At resonance, it's
really neat to see stuff fly apart -- but PCB's weren't the problem.
Everything else was.  The case, the mountings, the connectors.  Any
large items on the PCB are also at risk -- e.g. relays, transformers,
pots, switches, etc.  The PCB itself is purty damm tuff!!

So: (in the order I think most important, i.e. it's debatable)

- Protect your components from temperature extremes.  Put heatsinks
on things that get hot, but remember, that if the heat goes into a
closed box, it will just heat everything else up, and the heatsink
cannot work very well in a closed space.  Try to get the heat outside,
if at all possible, by providing aluminum or copper (low thermal
resistance) paths directly from the part to the case.  Your enclosure
should be aluminum.  Don't use steel or plastic as the heat can't get out.

- Protect your circuit from dirt and humidity.  Digital circuits
aren't too sensitive to current leakage due to moisture / dirt
combo but your sensors are analog and Hi-Z.  That means they are
seriously affected by even slight moisture or dirt.  (Frankly, I'm
amazed the EGO works as well as it does, since its connector is near
a serious heat source and is exposed to lots of dirt and water.)
It's not a bad idea to coat the PCB.  A solder mask is a real help
in preventing humidity problems (on the other hand, it can be a real
PITA if a break should occur in a trace somewhere).

- Protect it from undue vibration (don't put heavy things on the
PCB, or if you must, be sure the board is supported near them).

- Dont put stress on connectors.  Keep them away from dirt and
liquids (water, oil, grease, battery acid, etc.)

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

>Before being over-concerned about the chip operating temp,
>I'd look at the PC board itself.  These boards are nothing but
>thin etched metal overlaid on a fiber-glass like material with
>a lot of resin.  Every temperature cycle dries out a little more
>resin.  They get brittle with age.  They crack with vibration. They
>stress with each cycle - the board grows and shrinks at a rate
>different than the metal traces.  Think I'm kidding. A desk top
>pc generally lasts until the owner can't stand it anymore, while
>a laptop is generally a tosser at about 3 to 4 years.  Laptops
>get real brittle, with hairline cracks that can't be found or fixed.
>Symptoms like random freeze ups etc. = bad board. Why -
>vibration - temperature cycles - pressure cycles - that the 
>desktop living at human comfort temperatures and no motion
>stress does not see.
>
>If it was up to me (and most German, Japanese and American car 
>manufactures seem to follow the practice) I'd mount the board 
>where there was the least thermal and mechanical stress possible
>and minimize the problem.  Just a thought.
>